The study followed the method described in the published protocol and is in accordance with the CONSORT reporting guidelines [56, 57].

Research ethical approval

The research was reviewed by the NHS West Midlands—Edgbaston Research Ethics Committee with a favourable opinion (Reference: 22/WM/0178), and Health Research Authority approval was received (both dated 27th September 2022).

Study design

This feasibility study was a single-blind randomised controlled trial (RCT) with two arms: EMDR (plus standard neuropsychiatric care (NPC)) and standard NPC only. The two groups were compared at baseline, 3 months, 6 months, and 9 months. We recruited fifty adult patients with a diagnosis of FND (confirmed by a neurologist according to standardised diagnostic criteria), from a single site (neuropsychiatry service in the UK). After participants had completed their intervention period, a selection of them were invited to take part in semi-structured interviews regarding their experiences of the trial. Participants were invited to take part in interviews if they had completed their time in the trial and if they had been randomised to EMDR + NPC. Additionally, participants randomised to NPC were invited to be interviewed, once they had completed their time in the trial, and they were selected using purposive sampling to ensure a balance of symptom presentations. The trial therapists were also interviewed.

Allocation and blinding

Participants were randomised into EMDR + NPC or NPC alone in a 1:1 ratio. A stratified block randomisation (using randomly permuted blocks of sizes 2 and 4) was used so that there were similar numbers of participants with and without PTSD symptoms in each arm (PTSD diagnosis determined by diagnostic algorithm of the International Trauma Questionnaire). Randomisations were carried out by the Trial Manager (SV) using the randomisation function on REDCap.

The research assistant and statistician remained blind to treatment allocation.

Public and patient involvement (PPI)

PPI has been present in the research design process, throughout the delivery of the trial (PPI representatives within the Trial Management Group and Trial Steering Committee), and in contributions to trial materials, interview schedules, and publications.

Recruitment

Potential participants who expressed an interest in taking part in FND research were referred to the research team by the Neuropsychiatry service. The research assistant gave potential participants a summary of the study, the Participant Information Sheet and Informed Consent Form to review. Those willing to participate gave informed consent and were scheduled for a screening interview.

The screening interview was used to establish eligibility in terms of inclusion criterion (5) Reporting at least one traumatic event on the International Trauma Exposure Measure (ITEM), and exclusion criterion (7) Diagnosis of dissociative identity disorder or score in clinical range on “identity disturbance” subscale of Multiscale Dissociation Inventory (MDI). Potential participants needed to score in the non-clinical range of the subscale “Identity Disturbance” on the MDI to take part (score < 15). Eligible participants completed additional baseline measures, and provided demographic information, medical history, and previous psychological therapies attended.

Incentives

Participants were offered reimbursement for research-related travel costs (up to £20 per appointment). A non-contingent £25 incentive was offered to participants 9 months after informed consent, unrelated to trial completion. All participants who completed an interview were offered a £20 incentive.

Eligibility criteria

Inclusion criteria were (1) predominant diagnosis of functional seizures and/or functional motor symptoms, with diagnosis confirmed by neurologist; (2) aged 18 years or over; (3) capacity to consent; (4) willingness to attend regular psychological therapy sessions; (5) reporting at least one traumatic event on the International Trauma Exposure Measure (ITEM).

Exclusion criteria were (1) non-english speaking; (2) current ongoing adversity that is likely to interfere with psychological therapy, e.g. domestic violence, homelessness, unresolved compensation claim/litigation; (3) predominant diagnosis of borderline personality disorder*; (4) predominant diagnosis of chronic pain condition*, e.g. fibromyalgia; (5) predominant diagnosis of chronic fatigue syndrome*; (6) diagnosis of a psychotic disorder; (7) diagnosis of dissociative identity disorder or score in clinical range on “identity disturbance” subscale of MDI; (8) uncontrolled epileptic seizures; (9) diagnosis of an eating disorder; (10) current severe self harm or strong suicidal ideation that requires secondary care mental health services input; (11) current alcohol or drug harmful use or dependence; (12) current diazepam use exceeding the equivalent of 10 mg per day; (13) currently attending individual psychological therapy focused on FND or other specialist FND-specific treatment such as inpatient/outpatient multi-disciplinary treatment or intensive FND-specific physiotherapy.

*Comorbid diagnosis was acceptable, as long as FND was the predominant difficulty.

Study interventions EMDR plus standard neuropsychiatric care (EMDR + NPC)

Participants randomised to EMDR + NPC were offered up to sixteen EMDR sessions, (completed within six months), a one-month optional follow-up session (not counted as an EMDR session), and standard outpatient neuropsychiatric appointments (NPC). Participants could choose to attend 60–90 min sessions either in-person or virtually via MS Teams. A minimum of 8 sessions counted as completed treatment.

The standard EMDR protocol was delivered, but was adapted for FND presentations, following a treatment protocol developed by SC. Treatment has three stages: 1. Assessment, psychoeducation, target selection, and preparation for processing; 2. Processing of targets; 3. Ending of therapy. The initial sessions included education on FND, anxiety and dissociation; as well as formulating collaboratively with the participant regarding their FND symptoms’ development and maintenance. With the participant, target memories/images were chosen, such as (1) distressing memories associated with the time when symptoms began or FND symptoms generally; (2) distressing memories from past events that may be relevant to their FND symptoms; (3) FND symptoms themselves, when present in session, or an image of them; (4) distressing images about the future, e.g. image of having FND symptoms in front of others.

Standard neuropsychiatric care (NPC)

NPC was standard care and included 1–3 routine out-patient appointments with a neuropsychiatrist during the trial period. Participants were still invited to psycho-educational group interventions focused on FND routinely administered by the service. Their neuropsychiatrist referred for psychological therapy outside of the service for any comorbid conditions, but not for FND-specific psychological therapy or other therapies.

Training, supervision and fidelity checks

There were two trial therapists who had both completed EMDR-Europe Accredited Basic EMDR training (Parts 1–3): one trial therapist was a clinical psychologist who had completed their EMDR training in July 2021 (around 18 months post-training experience before trial). The other trial therapist was a cognitive–behavioural therapist who had completed EMDR training in April 2019 (around 44 months post-training experience before trial). Neither were accredited EMDR therapists yet, so can be considered relatively novice EMDR therapists. They both attended training on the FND-specific EMDR protocol. They received clinical supervision from SC (clinical psychologist and accredited EMDR Therapist), as well as external EMDR supervision from a clinical psychologist and EMDR Consultant (MC), with both supervisions every 2–4 weeks. Therapists completed a session record form after every session.

All sessions of EMDR were video-recorded (with the participants’ consent) and excerpts were shown in EMDR supervision. Fifteen randomly selected recordings of processing sessions were rated for fidelity by independent EMDR Consultants, using the EMDR Fidelity Rating Scale Version 2–Adverse Life Experiences Processing subscale [58].

Primary objectives and outcome measures

A mixed methods approach was used to establish feasibility and acceptability. The feasibility criteria were recruitment rate (% potentially eligible participants attending screening interview), intervention adherence (% participants randomised to EMDR + NPC who complete therapy), and outcome measure completion (% participants who complete outcome measures at all time points), with a target of above 70%, and a cut-off of above 50% demonstrating feasibility. If any criteria fell into the 50–70% bracket, ways to improve outcome(s) were considered [57].

Assessment of safety (adverse/serious adverse events) was examined between the two arms. Therapy satisfaction, choice of therapy session format (in-person/virtual), and therapy fidelity were No formal hypothesis testing of the feasibility datae evaluated and required sample size calculated to consider a primary outcome for a substantive RCT.

The nested qualitative study using interviews from participants and trial therapists will inform the substantive study. Participants’ qualitative experiences of the trial will be reported in another paper.

Outcome measures

Ecological Momentary Assessment using the m-Path App was used to measure FND symptoms [59]. This measurement approach was chosen as there is currently no single outcome measure that measures the variety and variability of FND symptoms [60]. Participants chose a maximum of two symptoms at the beginning of the trial period to rate, e.g. seizures, tremor, limb weakness, tingling/numbness, gait disturbance. They rated their chosen symptom(s) daily for two weeks at Baseline, and 3 months, 6 months, and 9 months follow-ups. They answered five questions each day in reference to their chosen symptom(s) (frequency, severity, interference, associated distress, associated preoccupation).

Other outcome measures were World Health Organisation Disability Assessment Schedule (WHODAS 2.0), EQ-5D-5L, PHQ-9, GAD-7, International Trauma Exposure Measure (ITEM), International Trauma Questionnaire (ITQ), Multiscale Dissociation Inventory (MDI), Adult Service Use Schedule (AD-SUS), Clinical Global Impression – Improvement Scale (CGI-I) rated by participant (single item, 7-point scale), and CGI-I rated by person nominated by participant (e.g. family member, partner, close friend, carer) (single item, 7-point scale) [61,62,63,64,65,66,67,68,69]. All the previously listed measures were completed at Baseline, 3 months, 6 months, and 9 months, except for the ITEM (Baseline only), the AD-SUS (Baseline and 9 months only), and the CGI-I (9 months only). Participants were also asked questions regarding ‘Agreement with diagnosis of FND’ (single item, 11-point scale), ‘Preference regarding treatment’ (EMDR + NPC, NPC or no preference), and ‘Belief in having been given the right treatment’ (single item, 11-point scale) at Baseline and 9 months. The satisfaction rating of treatment (single-item, 11-point scale) was asked at 9 months follow-up. Detailed descriptions of outcome measures and the schedule of assessments for participants can be found in the protocol paper [57].

Statistical analysis

This was a feasibility trial, and as such, a power calculation was not performed. Rather, a target recruitment of 50 participants was considered sufficient to examine the feasibility of a subsequent definitive RCT as a function of recruitment, adherence and retention rates [70]. A baseline table compared the demographic and key clinical characteristics between the two trial arms in a descriptive manner, summarising data by frequency (%) for categorical variables and mean (SD) for continuous variables that followed a Gaussian distribution and median (inter-quartile range) otherwise. Feasibility outcomes were summarised using descriptive statistics and compared to full-trial progression criteria, and acceptability of the intervention was evaluated primarily via examination of satisfaction of care, treatment preference and clinical global impression data at 9-month follow-up. Descriptive assessments of EMDR session characteristics (e.g., number/length of sessions, fidelity to EMDR protocol) and health care resource use stratified by treatment arm were also considered.

Analyses conducted to explore potential effects of EMDR therapy at 3-, 6- and 9-month follow-ups were, in the first instance, based on intention-to-treat (ITT) principles: all randomly assigned patients enrolled in the MODIFI trial who completed baseline measures were included. Each dependent variable was analysed by fitting Generalized Linear Mixed Models (GLMM), which allow for different numbers of repeated measurements between participants and response variables from different distributions, accounts for within-person clustering of data, and is appropriate for evaluation of intensive longitudinal data such as Ecological Momentary Assessment (EMA) [71, 72]. For individuals reporting functional seizures as one of their two symptoms, GLMM with linear (with identity link) and binomial (with logit link) distributions were respectively used to evaluate the impact of the intervention on EMA seizure frequency (transformed using Box–Cox methods to reduce positive skew) and on (the proportion of) days with at least one seizure; these models included condition (EMDR + NPC and NPC), time (baseline, 3-, 6-, and 9-month follow-ups) and condition*time (interaction) as fixed effects, and random intercepts at symptom (1 and 2), time, day (1–14), and person (subject-specific) levels. To assess treatment-specific changes in FND symptom severity, interference, distress and preoccupation (across all participants), four separate GLMM with linear distribution were administered; these included condition, time, symptom type (functional seizure, functional motor, and functional cognitive), symptom (1 and 2) and condition*time and condition*time*symptom type (interactions) as fixed effects and (as above) random intercepts at symptom (1 and 2), time, day (1–14), and person (subject-specific) levels. Inclusion of the condition*time*symptom interaction term in each model allowed the evaluation of any potential differential effects of the trial intervention according to (broad) FND symptom class (in addition to gauging an overall effect). GLMM with linear distribution was also employed to examine signal of efficacy for constructs assessed at each time point by standardised questionnaires (e.g., PTSD symptoms, disability levels, HRQoL and anxiety); these random intercept models included condition, time and condition*time interaction as fixed effects. Exploratory analyses (separately) considered potentially differential patterns of change in participants according to the presence of PTSD at pre-treatment and (for standardised questionnaire outcomes) whether the participants experienced functional seizures (by adding main effect and corresponding interactive fixed effects to GLMM).

All GLMM used restricted maximum likelihood (REML) estimation, which produces unbiased estimates in case of small sample sizes and employed first-order autoregressive (or first-order autoregressive moving average) structures for the residuals to account for time dependencies between adjacent assessments [73]. Where distribution of model residuals significantly differed from normality, dependent measures were transformed to better approximate a Gaussian distribution (e.g., EQ-5D-5L total value). Missing data were assumed to be missing at random (the least restrictive assumption); findings from Little’s Test of Missing Completely at Random supported this (for all tests, p > 127) [74].

Subsequent ‘per-protocol’ analyses, considering only those participants in the EMDR + NPC group completing treatment (i.e., attending ≥ 8 sessions) and with data at the time-point for the relevant measure (for EMA, this necessitated responses for no less than 5 of the 14 days in the time period) were administered; analysis of EMA data were performed in the manner above (i.e., fitting GLMM), while for standardised questionnaire outcome measures, between-group differences for primary and secondary outcome measures at each follow-up time-point, adjusted for pre-treatment score on the measure of interest, were calculated using analysis of covariance (ANCOVA), which relies on complete-case analysis).

Estimates of potential intervention effects (Hedge’s g) were calculated in GLMM by dividing the difference in change between the two trial arms (from estimated marginal means) at all post-randomisation time points by the corresponding pre-treatment pooled SD [75], and in per-protocol analyses by dividing the difference between (adjusted) mean scores at the time point by its pooled SD. An effect size of 0.2 was considered a small effect, 0.5 as a moderate effect and 0.8 as a large effect [76]. Interval estimates of these effects were produced in the form of 95% confidence intervals (CI), to explore imprecision and ensure the effect size subsequently chosen for powering a definitive trial is plausible [77]. No formal hypothesis testing of the feasibility data was undertaken.

Statistical analyses were undertaken using Statistical Package for the Social Sciences (SPSS) Version 29 (IBM Corporation, 2022), supplemented where required by Stata SE Version 16.0.